O(n) Layout-Coloring for Multiple-Patterning Lithography
and Conflict-Removal Using Compaction
Rani S. Ghaida*, Kanak B. Agarwal†, Sani R. Nassif†, Xin Yuan‡,
Lars W. Liebmann§, Puneet Gupta*
EE Dept, *
Austin Research Lab†
San Jose‡
Semiconductor R&D Center§
Multiple-Patterning Lithography
• Delays in EUV MP is inevitable for sub 20nm tech
– DP/TP in LELE, SADP
• Biggest challenge is coloring conflicts
• E.g., DP in LELE process
Rani S. Ghaida
2
C. Mack, IEEE Spectrum 08
Stitch
Native conflict needs layout change
Overview of the Framework
• Fast linear time coloring
• LP-based compaction for conflict removal
– Simultaneously fixes all conflicts without creating new conflicts
Rani S. Ghaida
3
Design with Conventional
Rules
Layout Simplification
(optional)
Coloring w/ Least Conflicts
Sign-off Conflicts LP for Conflict
Removal w/ Fixed Area
No conflicts
LP for Conflict Removal w/
Area Increase (optional)
MP Layers & Constraint Definition
Option
Option
Preferred Coloring
• Coloring of native conflicts affects efficiency of conflict removal
• Give preference for opposite coloring for certain violations over
others label violations critical vs. less-critical
– E.g., horizontal spacing violation more critical than vertical or
diagonal in case of vertical poly orientation
Rani S. Ghaida
4
Bad coloring Good coloring
M1 1st exposure M1 2nd exposure
Native conflict
Poly
Active
Contact
A
C C
A B B
O(n) Coloring
Rani S. Ghaida
5
Violating coloring matters
Non-violating coloring doesn’t
Stitches non-viol b/w violating
Projection
C
D
H
I J
E G
F
K
L
B
A Sub-comp. 1
Sub-comp. 2
Conflict
Graph
230X faster
than ILP coloring
Triple Patterning – Extending 2-Coloring to 3-Coloring
• Common 2-coloring cannot be extended to 3-coloring
– 3-coloring stitches can be almost anywhere!
Rani S. Ghaida
6
Projection all parts violating no stitches
No 3-Coloring Solution!
Valid coloring with TP stitch Valid coloring not possible
Unseen candidate stitch
7
TP Coloring Example
• Leverage TP Stitch Capability Stitch at S2S violating parts
• Color violating parts w/ C0/C1-C0/C2-C1/C2 coloring cycle
– Use existing infrastructure of DP coloring
• Works well but not for complex layouts simplifications needed
Rani S. Ghaida
Conflict-Removal Using Compaction
• Color define DRs between DP layers (e.g., M1A/M1B)
– Same-color spacing, ≠ color spacing, M1A/M1B overlap
– Overhang rules with top/bottom layers (union M1A M1B)
• Compaction Full legalization across all layers concurrently
Rani S. Ghaida
8
Design with Conventional
Rules
Layout Simplification
(optional)
Coloring w/ Least Conflicts
Sign-off Conflicts LP for Conflict
Removal w/ Fixed Area
No conflicts
LP for Conflict Removal w/
Area Increase (optional)
MP Layers & Constraint Definition
Option
Option
Sacrificing Unnecessary Layout Features
Rani S. Ghaida
9
Design with Conventional
Rules
Layout Simplification
(optional)
Coloring w/ Least Conflicts
Sign-off Conflicts LP for Conflict
Removal w/ Fixed Area
No conflicts
LP for Conflict Removal w/
Area Increase (optional)
MP Layers & Constraint Definition
Option
Option
Pin Segments Redundant Contacts
Recommended rule
Conflict Removal Results
• DP-compatible cells
– No area overhead for simple cells
– Modest area overhead (at most 9%) for complex cells and macros
– Few sacrificed redundant contacts (CA)
• Less than 1 min in real time for largest macro (460 trans.)
Rani S. Ghaida
10
No conflicts, 6.2% area increase
Original 5 conflicts
Same area, 2 conflicts
Effects of Sacrifice and Preferred Coloring
• Need both enhancement methods
• If enhancements not applied 2X more conflicts in final layout
Rani S. Ghaida
11
With Fixed Area
12
Final Notes
• Problems with Newly Created Tips
– One way use pessimistic projection non-optimal
– Less of a problem when using compaction-based legalization
• Methodology applicable for SADP, only need
– A layout-coloring method
– A set of design rules for SADP-compatible layout
Conflicts from
new tips
Thank you
Questions during poster session
Rani S. Ghaida
13